Currently the password generator notation '[…]' ignores a specified character weighting. For example:
[d\5\5\5\5\5\5\5\5\5\5] is the same as [d]
[\a\5\5\5\5\5\5\5\5\5\5] is the same as [\a\5]
[lddd] is the same as [ld]
[l\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!]{3}l{9} is the same as [l\!]{3}l{9}
Weighting character sets is a method of tuning the frequency of special characters in generated passwords that is distinct from the option to randomly permute password characters.
wellread1
2014-04-07
I should read the documentation more carefully. I see that the proposal has been considered, discarded and documented: http://keepass.info/help/base/pwgenerator.html#charset.
wellread1
2014-04-08
I believe the explanation for why it is not appropriate to weight (assign selection probabilities to) individual characters in the Password Generator documentation is a bit misleading. The statement below about sets is true, but it is irrelevant.
In mathematical terms, character sets are sets, not vectors. This means that characters can not be added twice to the set. Either a character is in the set or it is not.
The […] notation includes character repetition that is redundant in set notation. However, if this information is viewed as representing character weighting, it can be used easily to calculate the probability of randomly selecting a particular character from a set of characters having unequal probabilities of being selected. The distinction between a weighted set and an unweighted set disappears (except from the point of view of computational complexity) when one considers that the probability of randomly selecting a particular character from a character set of size N equal weighted characters is {1/N} for every member of the set.